[{"data":1,"prerenderedAt":95},["ShallowReactive",2],{"profile-255db4156af2c484002d80454c8c2b7d":3,"works-255db4156af2c484002d80454c8c2b7d":22,"scholar-stats":73},{"id":4,"name":5,"avatar":6,"org":7,"title":8,"bio":9,"intro":10,"badgeNo":11,"scholarUrl":11,"emailPublic":12,"github":13,"zhihu":11,"xhs":11,"experience":14,"invitedAt":15,"status":16,"directions":17,"works":21},"255db4156af2c484002d80454c8c2b7d","余元玺","https:\u002F\u002F6461-dayan-ai4s-d2gr3v6bc05b0eccd-1304768804.tcb.qcloud.la\u002Fweda-uploader\u002Ffffa15d8d4cb1bb44b8c37cfbf302951-1_%E4%BD%99%E5%85%83%E7%8E%BA.jpg","上海交通大学、上海灵肽科技有限公司","博士后、创始人","我的研究聚焦于结合PLM、小样本学习与实验验证，实现的蛋白质设计与改造。","余元玺，上海交通大学物理与天文学院致远荣誉博士。博士导师为洪亮教授。现为上海交通学博士后与上海灵肽科技创始人，主要研究方向为人工智能辅助的蛋白质设计，为Venus系列蛋白质模型的开发团队核心成员。","","y445980780@sjtu.edu.cn","https:\u002F\u002Fphy-yuanxi.github.io\u002F","2026-至今： 上海灵肽科技 创始人\n2025-至今：上海交通大学 博士后\n2020–2025：上海交通大学，致远荣誉博士\n2016–2020：上海交通大学，学士","2026.11.10","on",[18,19,20],"人工智能辅助的蛋白质定向进化","蛋白质语言模型","自动化实验",[],[23,37,49,61],{"id":24,"profileId":4,"type":25,"title":26,"authors":27,"venue":28,"date":29,"tags":30,"intro":34,"image":11,"url":35,"sort":36,"status":16},"40af8c4e6af2c748002d40f245e21307","paper","Deep learning-guided dual-fitness evolution of T7 RNA polymerase for enhanced stability and activity","姜帆、康里奇、李明辰、钟博子韬、陈晓霞、吴邦昊、李梦荣、余元玺、洪亮","Nucleic Acids Research, 54(6), gkag259","2026.03",[31,19,32,33],"蛋白质设计"," 酶工程","多目标优化","\u003Cp>\u003Cspan style=\"font-family:Source Han Sans SC, 思源黑体, sans-serif\">\u003Cspan style=\"font-size:14px\">\u003Cspan style=\"color:#344054\">将预训练蛋白质语言模型与迭代实验验证相结合，建立兼顾热稳定性与高温催化活性的双目标进化流程。通过单点突变筛选及多位点组合优化，获得性能改善的 T7 RNA 聚合酶，为以较少实验数据开展多目标酶改造提供方法。\u003C\u002Fspan>\u003C\u002Fspan>\u003C\u002Fspan>\u003C\u002Fp>","https:\u002F\u002Fdoi.org\u002F10.1093\u002Fnar\u002Fgkag259",1,{"id":38,"profileId":4,"type":25,"title":39,"authors":40,"venue":41,"date":42,"tags":43,"intro":46,"image":11,"url":47,"sort":48,"status":16},"9dd802d16af2c7ac003909be5368a2df","Venus-MAXWELL: Efficient Learning of Protein-Mutation Stability Landscapes using Protein Language Models","余元玺、姜帆、马新柱、张良、钟博子韬、欧阳万里、范贵生、虞慧群、洪亮、李明辰","Advances in Neural Information Processing Systems 38 (NeurIPS 2025)","2025年",[31,19,44,45]," 突变稳定性预测","深度学习","\u003Cp>\u003Cspan style=\"font-family:Source Han Sans SC, 思源黑体, sans-serif\">\u003Cspan style=\"font-size:14px\">\u003Cspan style=\"color:#344054\">提出 Venus-MAXWELL，将蛋白质突变稳定性预测转化为序列到稳定性景观的学习任务。利用蛋白质语言模型的预训练知识，并构建控制数据泄漏与冗余的评测集，在提高预测效率的同时支持多种语言模型。训练代码、模型权重和数据集已开源。\u003C\u002Fspan>\u003C\u002Fspan>\u003C\u002Fspan>\u003C\u002Fp>","https:\u002F\u002Fdoi.org\u002F10.52202\u002F085713-4581",2,{"id":50,"profileId":4,"type":25,"title":51,"authors":52,"venue":53,"date":54,"tags":55,"intro":58,"image":11,"url":59,"sort":60,"status":16},"9dd802d16af2c898003926816a98a5bf","A general temperature-guided language model to design proteins of enhanced stability and activity","姜帆、李明辰、董家君、余元玺、孙鑫宇、吴邦昊、黄晋、康里奇、裴羽丰、张良、王少杰、欧阳万里、范贵生、郑力荣、谭扬、胡志强、熊毅、冯雁、杨广宇、刘倩、宋杰、刘佳、洪亮、谈攀","Science Advances, 10(48), eadr2641","2024.11",[31,19,56,57],"零样本学习","稳定性与活性优化","\u003Cp>\u003Cspan style=\"font-family:Source Han Sans SC, 思源黑体, sans-serif\">\u003Cspan style=\"font-size:14px\">\u003Cspan style=\"color:#344054\">提出温度引导的蛋白质语言模型 PRIME，在没有目标蛋白突变实验数据的条件下推荐稳定性与活性改善的突变体。研究在 283 项蛋白质测定数据上开展评估，并对五种蛋白质进行实验验证，展示了人工智能辅助蛋白质设计的应用潜力。\u003C\u002Fspan>\u003C\u002Fspan>\u003C\u002Fspan>\u003C\u002Fp>","https:\u002F\u002Fdoi.org\u002F10.1126\u002Fsciadv.adr2641",3,{"id":62,"profileId":4,"type":25,"title":63,"authors":64,"venue":65,"date":66,"tags":67,"intro":70,"image":11,"url":71,"sort":72,"status":16},"40af8c4e6af2c8f1002db86f3e8440f8","Enhancing efficiency of protein language models with minimal wet-lab data through few-shot learning","周子宜、张良、余元玺、吴邦昊、李明辰、洪亮、谈攀","Nature Communications, 15, 5566","2024.07",[31,68,19,69],"小样本学习","蛋白质适应度预测","\u003Cp>\u003Cspan style=\"font-family:Source Han Sans SC, 思源黑体, sans-serif\">\u003Cspan style=\"font-size:14px\">\u003Cspan style=\"color:#344054\">提出 FSFP 小样本训练策略，将元迁移学习、排序学习与参数高效微调相结合，仅使用目标蛋白数十个带实验标签的单点突变体，即可改善蛋白质语言模型的适应度预测能力。研究在 87 个深度突变扫描数据集上进行评估，并通过 Phi29 DNA 聚合酶改造实验验证其在少量湿实验数据条件下指导蛋白质工程的有效性。\u003C\u002Fspan>\u003C\u002Fspan>\u003C\u002Fspan>\u003C\u002Fp>","https:\u002F\u002Fdoi.org\u002F10.1038\u002Fs41467-024-49798-6",4,{"views":74,"hotIds":91},{"853d22036a97e9a3017cb879318fbfa9":75,"a8add1456a96c24d012f36ee3a3e6e80":76,"a8add1456a96c24d012f36ef55b51133":77,"a8add1456a96c24d012f36f02ef5e9b4":78,"a8add1456a96c24d012f36f14e4da6e6":78,"a8add1456a96c24d012f36f2235a9643":79,"a8add1456a96c24d012f36f364c51aca":78,"a8add1456a96c24d012f36f43c5b2575":80,"a8add1456a96c24d012f36f5055a7687":81,"a2453f646ad864b9048cd5ea3c0a4cdf":82,"a2453f646ad864b9048cd5eb700f1c13":83,"572942446ad98b2501b418a57fd29780":84,"572942446ad98b2501b418a968974e47":85,"572942446ad98b2501b418a708985667":60,"572942446ad98b2501b418a866ceb5e8":72,"572942446ad9cf9201b83d000821e9dc":86,"572942446ad9cf9201b83cfe123c7162":87,"572942446ad9cf9201b83cff5362a431":88,"fa564f266ae1e2390042b77e034d91fd":89,"255db4156af2c484002d80454c8c2b7d":90,"255db4156af417090039154a0ff49242":36},106,92,31,13,11,12,22,10,57,98,88,46,36,51,38,16,[92,93,94],"572942446ad98b2501b418a968974e47","572942446ad98b2501b418a57fd29780","fa564f266ae1e2390042b77e034d91fd",1794993163112]